[luma]

Well, you learn from your mistakes and my mistake was to rely too much on the Travor split image focusing screen. I have purchased this screen a while ago and it seemed to work very well, especially with my manual focus lenses on my Pentax KX.

Shortly after installing the Travor screen, I decided to fine tune the autofocus on my camera using the DEBUG menu. I used the focusing chart where you shoot at 45 degree focusing the camera on the bar in the center. I used the kit 18-55mm lens in spot focus mode, pointing the camera in such a way that the center bar was right in the middle of my split image circle.
All went OK, except the tests indicated that my camera had a strong front focusing that required -140 points to correct, way too high, but I took it for granted. I use manual focus lenses quite a bit so I did not pay too much attention to IQ from my autofocus lenses.

Then I have recently purchased the DA 10-17mm FE lens and the initial results were just terrible in terms of image sharpness, some pictures we outright blurred. Then it occurred to me, let’s try the autofocus test with the FE lens and sure enough, there was a strong back focus. To tune the AF for this lens, I had to go to -20 points from the original -140 points. I thought, something must be wrong here…

Then I got this “brain wave” and decided to check if the center of the split image circle is identical with the center of the LiveView, which is what the image sensor is seeing. Sure enough, the actual image centre where the camera is spot focusing was at the bottom of my split image circle! So basically, when doing my AF tests, I was not focusing the camera on the center bar of the focus chart, but rather below biasing thus the AF correction.
So, now taking this into account and using the LiveView, I repeated my AF correction tests and both lenses indicate an optimum -20 points adjustment for the perfect autofocus. The best part is that my new 10-17mm FE is producing now perfectly sharp pictures.

A final comment – it seems that misadjusting the AF correction has much more effect on the wider lenses than on the longer lenses. While the incorrect -140 points adjustment was totally blurring the 10-17mm lens, it was mildly noticeable on the 18-55mm lens and not noticeable on the 50-200mm lens at all.

[xjx]

I just took a quick check of my camera with Travor screen. You are right, the split image circle is located higher than the actual image center. That explains the minor discrepancy between the split image and the focus confirmation when I take picture at an upward angle.

[timstone]

Wait....there's a debug menu on the K-x???

[luma]

Wait....there's a debug menu on the K-x???

There is many references to it on the web, here is one:
deejjjaaaa: Using Pentax Kx debug mode to do a focus adjustment

[Blue]

Has anyone had this problem on the K200d or K20d?

[Lowell Goudge]

This whole thread should not be about the focusing screen, but the falicy that diagonal focusing targets should be used for FF/BF adjustment.

When looking through the viewfinder, you do not know exactly what the focusing sensor is aimed at. As a result, and perhaps shown here as a worst case, what the OP thought was the focus point, was not, causing misadjustment of the focusing.

The only reliable focusing test is to focus on a target that is perpendicular to the lens axis, a sloped target is simply NOT the way to do things.

Perhaps the focusing screen is not centered, I can't say, but the fact is that the OP was not pointing the camera at what the sensor was considering. If he repeated the test on a flat surface, he would focus correctly.

As for the screen, being off center is a problem, in terms of using it on other AF lenses because the split image will lead you to focus at the center of the split, that is just human nature

[mattdm]

I recommend this focus chart:

Jeffrey Friedl's Blog Jeffrey’s Autofocus Test Chart

It's designed so that the "correct" focus bar is the only thing with enough contrast to grab the focus sensor, even if you're a little off on where you think it is.

[Parallax]

I have the classic focus chart mounted on the side of a cardboard box with a reference line drawn vertically that runs through the exact center of the chart. I put an X on the counter top with a pencil and the box is positioned so it is parallel to the sensor plane with the line at the bottom centered on the X. I align the camera so that the center of the chart is in the exact center of the viewfinder, focus on the chart and then rotate the chart to 45 degrees keeping the reference line centered on the x. This way I know that the camera focused at the exact distance to the center of the chart, even if the actual position of the focus point is not in the exact center of the image.

[bjake]

I have a k-x that the autofocus seems not to be correct.Haven't really check the kit lens that came with the camera.Put a fa 28-80 on the camera did a series of pictures with most all looking not sharp.Did an almost identical series using manual focus that were much better.Have looked at the debug article,not being computer or technical savvy didn't quite understand it.
Create a plain text/ascii file named "MODSET.492" with just one text line in the root folder of your SD flash card:

[OPEN_DEBUG_MENU]

NOTE: the line must ends with CR/LF (as it is a customary in "MS Windows/DOS" text files - two bytes : OxOD (13) = "CR" and OxOA (10) = "LF").
I am thinking one put a sd card in the computer and type 'MODSET.492' then put it in the camera? Should the sd card be formatted first?

The whole procedure has me confused.
Jake

[Lloydy]

This whole thread should not be about the focusing screen, but the falicy that diagonal focusing targets should be used for FF/BF adjustment.

When looking through the viewfinder, you do not know exactly what the focusing sensor is aimed at. As a result, and perhaps shown here as a worst case, what the OP thought was the focus point, was not, causing misadjustment of the focusing.

The only reliable focusing test is to focus on a target that is perpendicular to the lens axis, a sloped target is simply NOT the way to do things.

Perhaps the focusing screen is not centered, I can't say, but the fact is that the OP was not pointing the camera at what the sensor was considering. If he repeated the test on a flat surface, he would focus correctly.

As for the screen, being off center is a problem, in terms of using it on other AF lenses because the split image will lead you to focus at the center of the split, that is just human nature

Can you expand on this Lowell, I've used the sloping one before and I'm very interested to see what advantages the perpendicular method has, and how it's done ?

[mattdm]

Check this out for an interesting alternative method to traditional focus-test charts:

What is the best way to micro-adjust a lens? - Photography - Stack Exchange

[Parallax]

Can you expand on this Lowell, I've used the sloping one before and I'm very interested to see what advantages the perpendicular method has, and how it's done ?

The advantage of the perpendicular method is that you know exactly what distance the camera focused at because it had only one choice.

[kitkat]

If the viewfinder is in focus on a perpendicular target, and the recorded sensor image is out of focus, how do you tell if the focus is behind or in front of the perpendicular target? Isn't that the reason to use a sloping target?

[Parallax]

My method is to get focus with the target perpendicular, then turn the target 45 degrees (using a reference point to keep the center distance unchanged), then expose.

[Quicksand]

Another way to do it (if you have a camera that supports this) is to focus normally with the split prism until everything looks lined up and sharp, then switch on live view and manually turn the focus ring until the image is perfectly in-focus on the live view screen.

The direction you had to turn the focus ring to get perfect focus will tell you whether you're front-focused (if you turned toward infinity) or back-focused (if you turned away from infinity).

EDIT: The original way I wrote this was a good description of how to calibrate autofocus, which is not the problem here! Fixed.